Research On Semicompetiting Risk Model Based On MM Algorithm

In clinical medical research,patients may be susceptible to a variety of possible outcomes.When he encounters multiple events in subsequent processes,multiple invalidation time data is generated.A semicompeting risk framework in which subjects experience two different types of terminal events and nonterminal events,is considered in this thesis.And terminal event censors nonterminal event.This type of data is called semi-competing risk data.Random effects or “frailty”terms are often introduced to study dependence between event times to model such data.A illness-death model with shared frailty to study the dependence between events is established in this thesis.Structure,and covariate effect each event.A class of MM algorithm is adopted to estimate,whose main idea is to decompose the high-dimensional objective function into the form of each low-dimensional function sum,and then carry out simulation research on proposed method to evaluate effect.Finally,empirically analyze data of randomized clinical trials of nasopharyngeal carcinoma is provided.In simulation study,error and mean square error of each parameter are relatively small.It shows that effect of estimating illness-death model with MM algorithm is ideal,and error and mean square error of each parameter are reduced with increase of sample size.In empirical analysis,a positive correlation between recurrence and death is observed,which indicates that relapse accelerates onset of death.This means that in practice it is necessary to consider the correlation between non-terminal events and terminal events in order to achieve a more accurate estimate.